Repropack processes command output that may contain credentials or personal information. Its redaction layer reduces accidental disclosure; it does not make arbitrary logs safe by definition.

• All processing is local.
• Repropack does not upload reports or telemetry.
• .env files, source contents, Git diffs, credential directories, and shell history are not read.
• Sensitive environment values are held in process memory only to locate matching text in report fields.
• Saved reports contain aliases and redaction counts, not an alias-to-secret lookup table.
• Combined captured stdout/stderr is limited to 2 MiB by default.

The redactor is intended to catch common accidental disclosures in command arguments, logs, filenames, and collected metadata. It recognizes sensitive environment values, common authentication material, selected API-key formats, URL passwords, emails, and home-directory usernames.

It may not recognize:

• Application-specific credentials with unknown formats
• Partial or transformed secrets
• Secrets split across multiple output chunks or fields
• Sensitive business data that does not resemble a credential
• Identifiers embedded in binary or malformed text
• Values printed directly to the live terminal before report creation

Always inspect the generated report before attaching it to an issue or support ticket. For the safest inspection flow, use --preview, which suppresses raw live output and writes no files.

Do not include real credentials or private reports in a public issue. Once the repository is published, use its private vulnerability-reporting channel when available. Until then, report the issue directly to the maintainer with synthetic reproduction data.